Post #1854994

Running my tar parser on all the tar files in the Debian archive, to find cases I haven't handled yet. It's not done yet, but so far I've collected 2 num_read == sizeof s->buffer 3 symlink_size < LEN_LINK_TARGET 3243 unexpected character in octal 780 unhandled file type 17 unused tar header has data 8 value_len < sizeof s->filename Basically, I've plastered a bunch of checks all over the code that are like assert(), but remain in Release code. All of these are hard errors so far, and I will need to add more checks and also make the parser less strict in places where it is overly so. "num_read == sizeof s->buffer" means that the tar file is not a multiple of 512 bytes in size. "symlink_size < LEN_LINK_TARGET" means that the link target field is used completely and not NUL terminated. That's probably legal, but I'm amazed someone is running into this. Also, I suspect there is an encoding for even longer symlinks. "unexpected character in octal" seems to be mostly octal numbers that are right-justified in the field with spaces, which should be legal. "unhandled file type" refers to the file type marker inside the archive -- I handle directories, files, symlinks and extension headers for the whole archive and for individual files. So that's a TODO item. "unused tar header has data" happens when the first byte of the filename is NUL, but there are non-NUL bytes in that block. Need to investigate what these are. "value_len < sizeof s->filename" means that there is a long file name with more than 255 bytes. Getting the full statistics will take a while, and then I need to fix the program and re-run everything.